HomeCloudSoftwareHardware
DepthAI Components
DepthAI Tutorials
DepthAI Perception
Tools
ROS
Neural Networks
DepthAI API References

ON THIS PAGE

  • Thermal People Detection
  • Setup
  • Source code
  • Pipeline

Thermal People Detection

This example uses a thermal sensor, found on OAK Thermal camera, to detect people using a neural network. The detected people are then displayed on the thermal camera frames.

Setup

Please run the install script to download all required dependencies. Please note that this script must be ran from git context, so you have to download the depthai-python repository first and then run the script
Command Line
1git clone https://github.com/luxonis/depthai-python.git
2cd depthai-python/examples
3python3 install_requirements.py
For additional information, please follow the installation guide.

Source code

Python
C++

Python

Python
GitHub
1#!/usr/bin/env python3
2
3import depthai as dai
4import cv2
5from pathlib import Path
6import numpy as np
7import sys
8
9
10nnPath = str((Path(__file__).parent / Path('../models/yolov6n_thermal_people_256x192_openvino_2022.1_6shave.blob')).resolve().absolute())
11if len(sys.argv) > 1:
12    nnPath = sys.argv[1]
13
14if not Path(nnPath).exists():
15    import sys
16    raise FileNotFoundError(f'Required file/s not found, please run "{sys.executable} install_requirements.py"')
17
18labels = ["person"]
19
20device = dai.Device()
21
22pipeline = dai.Pipeline()
23nnet = pipeline.create(dai.node.YoloDetectionNetwork)
24nnet.setBlobPath(nnPath)
25nnet.setConfidenceThreshold(0.5)
26nnet.setNumClasses(1)
27nnet.setCoordinateSize(4)
28nnet.setIouThreshold(0.4)
29
30thermalCam = pipeline.create(dai.node.Camera)
31thermalCam.setBoardSocket(dai.CameraBoardSocket.CAM_E)
32thermalCam.setPreviewSize(256, 192)
33
34thermalCam.raw.link(nnet.input)
35
36rawOut = pipeline.createXLinkOut()
37rawOut.setStreamName("preview")
38thermalCam.preview.link(rawOut.input)
39
40xoutNn = pipeline.createXLinkOut()
41xoutNn.setStreamName("nn")
42nnet.out.link(xoutNn.input)
43
44xoutPass = pipeline.createXLinkOut()
45xoutPass.setStreamName("pass")
46nnet.passthrough.link(xoutPass.input)
47
48device.startPipeline(pipeline)
49
50qNn = device.getOutputQueue(name="nn", maxSize=2, blocking=False)
51qPass = device.getOutputQueue(name="pass", maxSize=2, blocking=False)
52qPreview = device.getOutputQueue(name="preview", maxSize=2, blocking=False)
53
54cv2.namedWindow("nnet", cv2.WINDOW_NORMAL)
55cv2.namedWindow("raw", cv2.WINDOW_NORMAL)
56cv2.resizeWindow("nnet", 640, 480)
57cv2.resizeWindow("raw", 640, 480)
58
59while True:
60    inNn = qNn.get()
61    inPass = qPass.tryGet()
62    inPreview = qPreview.get()
63    if inNn and inPass:
64        frame = inPass.getCvFrame().astype(np.float32)
65        min_, max_ = frame.min(), frame.max()
66        colormappedFrame = cv2.normalize(frame, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)
67        colormappedFrame = cv2.applyColorMap(colormappedFrame, cv2.COLORMAP_MAGMA)
68
69        detections = inNn.detections
70        for detection in detections:
71            xmin = max(0.0, detection.xmin)
72            ymin = max(0.0, detection.ymin)
73            xmax = min(1.0, detection.xmax)
74            ymax = min(1.0, detection.ymax)
75            pt1 = int(xmin * 256), int(ymin * 192)
76            pt2 = int(xmax * 256), int(ymax * 192)
77            cv2.rectangle(colormappedFrame, pt1, pt2, (0, 255, 0))
78            cv2.putText(colormappedFrame, labels[detection.label], pt1, cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2, cv2.LINE_AA)
79        cv2.imshow("nnet", colormappedFrame)
80    if inPreview:
81        cv2.imshow("raw", inPreview.getCvFrame())
82
83    if cv2.waitKey(1) == ord("q"):
84        break

Pipeline

React Flow

Need assistance?

Head over to Discussion Forum for technical support or any other questions you might have.